Aleksander / Taylor Artificial Neural Networks, 2

Proceedings of the 1992 International Conference on Artificial Neural Networks (ICANN-92) Brighton, United Kingdom, 4-7 September, 1992
1. Auflage 2014
ISBN: 978-1-4832-9806-1
Verlag: Elsevier Science & Techn.
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, 878 Seiten, Web PDF

This two-volume proceedings compilation is a selection of research papers presented at the ICANN-92. The scope of the volumes is interdisciplinary, ranging from the minutiae of VLSI hardware, to new discoveries in neurobiology, through to the workings of the human mind. USA and European research is well represented, including not only new thoughts from old masters but also a large number of first-time authors who are ensuring the continued development of the field.

Autoren/Hrsg.

Aleksander, I.

Taylor, J.

Inhaltsverzeichnis

1;Front Cover
;1
2;Artificialneural
Networks,2;4
3;Copyright Page
;5
4;Table of Contents;6
5;Part I: Neurobiology 1;18
5.1;Chapter 1. Temporal sequence storage;20
5.1.1;Abstract;20
5.1.2;1. INTRODUCTION;20
5.1.3;2. TEMPORAL NEURONS;21
5.1.4;3. PREDICTIVE SELF-LEARNING;22
5.1.5;4. TEMPORAL TOPOGRAPHIC MAPS;23
5.1.6;5. BIOLOGICAL REALISATIONS;24
5.1.7;6· REFERENCES;24
5.2;CHAPTER 2. GAMMA-BAND AND BETA-BAND CORTICAL OSCILLATIONS;26
5.2.1;ABSTRACT;26
5.2.2;INTRODUCTION;26
5.2.3;PROJECTION-MEDIATED OSCILLATIONS;26
5.2.4;A REVISION OF OUR EARLIER STUDY;27
5.2.5;THE POSSIBLE ROLE OF BETA-BAND AND GAMMA-BAND OSCILLATIONS;28
5.2.6;CONCLUSION;29
5.2.7;REFERENCES;29
5.3;Chapter 3. Synchronization and label-switching in networks of laterally
coupled model neurons;30
5.3.1;Abstract;30
5.3.2;1. INTRODUCTION;30
5.3.3;2. NETWORK MODEL AND MEASURE OF SYNCHRONIZATION;30
5.3.4;3. GLOBAL HOMOGENEOUS STIMULATION;31
5.3.5;4. SELECTIVE STIMULATION OF SUBPOPULATIONS;32
5.3.6;5. DISCUSSION;33
5.3.7;REFERENCES;33
5.4;Chapter 4. A model for the organisation of neocortical maps;34
5.4.1;Abstract;34
5.4.2;1 Introduction;34
5.4.3;2 The model;34
5.4.4;3 Conclusion;37
5.4.5;References;38
5.5;Chapter 5. The Neuronal Computation Time;40
5.5.1;Abstract;40
5.5.2;1. INTRODUCTION;40
5.5.3;2. RESULTS;41
5.5.4;3. DISCUSSION;42
5.5.5;4. CONCLUSION;43
5.5.6;5. REFERENCES;43
5.6;Chapter 6. Redundancy reduction of a Gabor representation: a possible computational role for feedback from primary visual cortex to
lateral geniculate nucleus;44
5.6.1;Abstract;44
5.6.2;1. PHYSIOLOGICAL BACKGROUND;44
5.6.3;2. ENCODING OF IMAGES WITH GABOR TRANSFORMS;44
5.6.4;4. CORTICO-THALAMIC IMPLEMENTATION;45
5.6.5;5. SIMULATIONS;46
5.6.6;6. IMPLICATIONS OF THE MODEL;47
5.6.7;Acknowledgements;47
5.6.8;7. REFERENCES;47
5.7;Chapter 7. Connection weights based on molecular mechanisms
in Aplysia neuron synapses;48
5.7.1;Abstract;48
5.7.2;1. Introduction;48
5.7.3;2. Model description;49
5.7.4;3. Results;51
5.7.5;4. Acknowledgements;51
5.7.6;5. References;51
5.8;CHAPTER
8. DYNAMIC PHENOMENA IN THE OLFACTORY BULB;52
5.8.1;Abstract;52
5.8.2;1 Introduction;52
5.8.3;2 The model;52
5.8.4;3 Simulation results;53
5.8.5;Acknowledgment;55
5.8.6;References;55
5.9;CHAPTER
9. DYNAMIC PHENOMENA IN THE OLFACTORY BULB;56
5.9.1;Abstract;56
5.9.2;1 Introduction;56
5.9.3;2 The model;56
5.9.4;3 Simulation results;58
5.9.5;4 Discussion;59
5.9.6;Acknowledgment;59
5.9.7;References;59
5.10;Chapter 10. A Model of Adaptive Development of Complex Cortical Cells;60
5.10.1;Abstract;60
5.10.2;1. INTRODUCTION;60
5.10.3;2. A BASIC MODEL OF COMPLEX CELL ADAPTATION;60
5.10.4;3. A MORE DETAILED MODEL;61
5.10.5;4. ACKNOWLEDGEMENTS;62
5.10.6;REFERENCES;62
5.11;Chapter 11. Activity-Induced "Colour Blob" Formation;64
5.11.1;Abstract;64
5.11.2;1. INTRODUCTION;64
5.11.3;2. PRINCIPAL COMPONENT ANALYSIS;64
5.11.4;3. A NETWORK THAT DEVELOPS "COLOUR BLOBS";65
5.11.5;4. RESULTS;66
5.11.6;5. ACKNOWLEDGEMENTS;66
5.11.7;REFERENCES;66
6;Part II: Neurobiology 2;68
6.1;Chapter 12. Computational analysis of the operation of a real neuronal network
in the brain: the role of the hippocampus in memory;70
6.1.1;Abstract;70
6.1.2;1. HIPPOCAMPAL FUNCTION;70
6.1.3;2. THE HIPPOCAMPAL SYSTEM;71
6.1.4;3. MEMORY CAPACITY OF THE CA3 NETWORK;73
6.1.5;4. THE ROLES OF THE INPUT SYSTEMS;75
6.1.6;REFERENCES;77
6.2;Chapter 13. Architectural Consequences of Mapping 3D Space Representations
onto 2D;78
6.2.1;Abstract;78
6.2.2;1. INTRODUCTION;78
6.2.3;2. SPACE REPRESENTATION IN A MOVING FRAME;78
6.2.4;3. BIOLOGICAL CONTRAINTS;79
6.2.5;4. EMBEDDING A 3D MECHANISM INTO A 2D ARCHITECTURE;79
6.2.6;5. SIMILARITIES WITH CEREBELLAR STRUCTURES?;81
6.2.7;6. CONCLUSION;81
6.2.8;7. REFERENCES;81
6.3;Chapter 14. A Neural Network for Fixation Point Selection based on Spatial Knowledge;82
6.3.1;Abstract;82
6.3.2;1 Introduction;82
6.3.3;2 Model;82
6.3.4;3 Computer simulation;84
6.3.5;4 Discussion;85
6.3.6;References;85
6.4;Chapter 15. Phase Transitions, Hysteresis and Overshoot
in Developing Neural Networks;86
6.4.1;Abstract;86
6.4.2;1 Introduction and Summary;86
6.4.3;2 The Model;87
6.4.4;3 Results;87
6.4.5;4 Conclusions and Discussion;89
6.4.6;5 References;89
6.5;Chapter 16. A Distributed Model of the Representational States in Classical Conditioning;90
6.5.1;Abstract;90
6.5.2;1. Introduction;90
6.5.3;2. Classical Conditioning;90
6.5.4;3. Classical Conditioning as Unsupervised Sequence Learning;91
6.5.5;4. Simulations;93
6.5.6;5. Conclusion;94
6.5.7;References;94
6.6;Chapter 17. A Neural Model of Cortical Cells Characterized by Gabor-like
Receptive Fields - Application to Texture Segmentation;96
6.6.1;Abstract;96
6.6.2;1. INTRODUCTION;96
6.6.3;2. MODELING INTRACORTICAL PROPERTIES;96
6.6.4;3. APPLICATION TO TEXTURE SEGMENTATION;98
6.6.5;4. REFERENCES;99
6.7;CHAPTER 18. STOCHASTIC AND OSCILLATORY BURST ACTIVITIES
IN A MODEL OF SPIKING NEURONS;100
6.7.1;Abstract;100
6.7.2;Introduction;100
6.7.3;Simulations and Results;101
6.7.4;References;103
6.8;Chapter 19. Reinforcement Learning: On Being Wise During the Event;104
6.8.1;Abstract;104
6.8.2;1. Introduction;104
6.8.3;2. Implementing Backchaining in a Neural Network;105
6.8.4;3. Simulations in a 2D Environment;106
6.8.5;4. Conclusion;107
6.8.6;References;107
6.9;Chapter 20. A Simulation of the Gated Thalamo-Cortical
Model;108
6.9.1;Abstract;108
6.9.2;1. INTRODUCTION;108
6.9.3;2. THE MODEL;108
6.9.4;3. RESULTS;109
6.9.5;REFERENCES;109
6.10;Chapter 21. The Biomagnetic Inverse Problem;112
6.10.1;Abstract;112
6.10.2;1. INTRODUCTION;112
6.10.3;2. EXPECTATION VALUES FOR THE CURRENT SOURCES;113
6.10.4;3. RECONSTRUCTION OF POINT SOURCES;114
6.10.5;4. REFERENCES;115
6.11;Chapter 22. Limitations of Logical Reasoning in Neural Networks
and Reasoning by Analogy;116
6.11.1;Abstract;116
6.11.2;Logic operations in neural networks;116
6.11.3;Biological
systems;119
6.11.4;Conclusion;119
6.11.5;References;119
7;Part III: Algorithms 1;120
7.1;Chapter 23. G-Nets and Learning Recurrent Random Networks;122
7.1.1;Abstract;122
7.1.2;1. INTRODUCTION;122
7.1.3;2. A LEARNING ALGORITHM;124
7.1.4;3. REFERENCES;125
7.2;Chapter 24. On a Class of Efficient Learning Algorithms for Multi-Layered Neural Networks;126
7.2.1;Abstract;126
7.2.2;1. BACKPROPAGATING THE DESIRED NET-OUTPUT;126
7.2.3;2. A NEW CLASS OF SUBNET ALGORITHMS;128
7.2.4;3. A TEST-EXAMPLES;129
7.2.5;4. REFERENCES;130
7.3;Chapter 25. A hybrid genetic algorithm for training neural networks;132
7.3.1;Abstract;132
7.3.2;1. INTRODUCTION;132
7.3.3;2. THE HYBRID ALGORITHM;133
7.3.4;3. SIMULATION RESULTS;134
7.3.5;4. CONCLUSIONS;134
7.3.6;5. REFERENCES;135
7.4;Chapter 26. Matching the topology of a neural net to a particualr problem: Preliminary results using correlation analysis as a pruning tool;136
7.4.1;Abstract;136
7.4.2;1. Inroduction;136
7.4.3;2. Proposed Methodology;136
7.4.4;3. Neural networks - learning;137
7.4.5;4. Results using correlation analysis;137
7.4.6;5. Conclusions;137
7.4.7;6. References;138
7.5;CHAPTER 27. PERFORMANCE COMPARISON OF LEARNING ALGORITHMS
IN HOPFIELD NETWORKS;140
7.5.1;Abstract;140
7.5.2;1. LEARNING IN HOPFIELD NETWORKS;140
7.5.3;2. COMPARISON FOR THE FULLY CONNECTED NETWORK;141
7.5.4;3. LEARNING ON LOCALLY CONNECTED ARCHITECTURES;142
7.5.5;4. REFERENCES;143
7.6;Chapter 28. A Study of Maximum Matching on Boltzmann Machines;144
7.6.1;Abstract;144
7.6.2;1 Introduction;144
7.6.3;2 The Boltzmann Machine Model;145
7.6.4;3 The Main Results;146
7.6.5;4 Conclusion;147
7.6.6;References;147
7.7;Chapter 29. Generation of Inhibitory Connections to Minimize
Internal and External Entropy;148
7.7.1;Abstract;148
7.7.2;1 Introduction;148
7.7.3;2 Learning Method with Complexity Term;149
7.7.4;3 Internal and External Entropy;150
7.7.5;4 Results;151
7.7.6;References;151
7.8;Chapter 30. The extended quickprop;152
7.8.1;Abstract;152
7.8.2;References;155
7.9;Chapter 31. Evolutionary construction algorithms for topology
conserving neural nets;158
7.9.1;Abstract;158
7.9.2;1 INTRODUCTION;158
7.9.3;2 A COEVOLVING NETWORK WITH LOCAL COMPETITION;158
7.9.4;3 T-NETWORKS;160
7.9.5;4 CONCLUSIONS AND OUTLOOK;161
7.9.6;References;161
7.10;Chapter 32. SCAWI: an algorithm for weight initialization of a
sigmoidal neural network;162
7.10.1;Abstract;162
7.10.2;1 Description of the method;162
7.10.3;2 Benchmarks Results;163
7.10.4;References;165
7.11;Chapter 33. An Approximation-Based Model of Associative Memory;166
7.11.1;Abstract;166
7.11.2;1. INTRODUCTION;166
7.11.3;2. REGULARIZATION THEORY;166
7.11.4;3. REGULARIZATION AND ASSOCIATIVE MEMORY;167
7.11.5;4. DISCUSSION;168
7.11.6;5. REFERENCES;169
7.12;Chapter 34. Pruning Neural Nets by Genetic Algorithm;170
7.12.1;Abstract;170
7.12.2;1. Introduction;170
7.12.3;2. Experiments;171
7.12.4;3. Conclusions;172
7.12.5;Acknowledgements;172
7.12.6;References;173
7.13;Chapter 35.
Optimizing Multilayer Networks Layer per Layer without Backpropagation;174
7.13.1;Abstract;174
7.13.2;1. Introduction;174
7.13.3;2. Presentation of the
criterion;175
7.13.4;3.
Methodology;176
7.13.5;4. Illustration of performance;176
7.13.6;5. Conclusion and further
work;177
7.13.7;References;177
7.14;Chapter 36. Entropy and Generalization in Feedforward Nets;178
7.14.1;abstract;178
7.14.2;1. Layers as Channels;178
7.14.3;2. Two Forms of Generalisation;180
7.14.4;References;181
7.15;Chapter 37. Improving Convergence of Back-Propagation by
Handling Flat-Spots in the Output Layer;182
7.15.1;Abstract;182
7.15.2;1 Introduction;182
7.15.3;2 Eliminating Flat-Spots in the Output Layer;183
7.15.4;3 Implementation Issues;184
7.15.5;4 Simulation Results;184
7.15.6;5 Discussion;186
7.15.7;6 Conclusion;187
7.15.8;Acknowledgement;188
7.15.9;References;188
7.16;Chapter 38. Fast Convergence of Neural Networks
by Application of a New Min-Max Algorithm;190
7.16.1;Abstract;190
7.16.2;1. INTRODUCTION;190
7.16.3;2. DESCRIPTION OF THE PROPOSED ALGORITHM;191
7.16.4;3. EXPERIMENTAL RESULTS;192
7.16.5;4. REFERENCES;192
8;Part IV: Algorithms 2;194
8.1;Chapter 39. The Munificence of High Dimensionality;196
8.1.1;Abstract;196
8.1.2;1 Introduction;196
8.1.3;2 The Geometry of Quasiorthogonal Sets;197
8.1.4;3 The Geometry of Feedforward Neural Network
Weight Spaces;200
8.1.5;References;209
8.2;Chapter 40. A Tripartite Framework for Artificial Neural Networks;210
8.2.1;Abstract;210
8.2.2;1. DESCRIBING AND CLASSIFYING NEURAL NETWORKS;210
8.2.3;2. A FRAMEWORK FOR ARTIFICIAL NEURAL NETWORKS (FANN);211
8.2.4;3. REFERENCES;213
8.2.5;Acknowledgment;213
8.3;Chapter 41. A Backpropagation Algorithm for Neural Networks Using Random Pulse
Streams;214
8.3.1;Abstract;214
8.3.2;1. Introduction;214
8.3.3;2. Derivation of a back-propagation algorithm;215
8.3.4;3. Simulation Results;216
8.3.5;4. Conclusions;217
8.3.6;5. References;217
8.4;Chapter 42. A collection of constraint design rules for neural optimization problems;218
8.4.1;Abstract;218
8.4.2;1. Introduction;218
8.4.3;2. Theory;218
8.4.4;3. Collection of design rules;219
8.4.5;4. Constraints converted into more than one k-out-of-n-constraint;220
8.4.6;5. Summary;221
8.4.7;6. References;221
8.5;Chapter 43. Optimization of the Rectilinear Steiner Tree
Using a Mean Field Theory Model;222
8.5.1;1. The Steiner Tree Problem;222
8.5.2;2. The Neural Net Model Used;222
8.5.3;3. Results;223
8.5.4;4. Summary;225
8.5.5;5. References;225
8.6;Chapter 44. A new learning scheme for dynamic self-adaptation
of learning-relevant parameters;226
8.6.1;Abstract;226
8.6.2;1 Introduction;226
8.6.3;2 Procedure;227
8.6.4;3 Convergence proof;227
8.6.5;4 Performance;227
8.6.6;5 Conclusion;228
8.6.7;Acknowledgements;229
8.6.8;References;229
8.7;Chapter 45. Growing Cell Structures –
a Self-organizing Network in k Dimensions;230
8.7.1;Abstract;230
8.7.2;1 Introduction;230
8.7.3;2 Network Architecture;231
8.7.4;3 Network Dynamics;231
8.7.5;4 Computational Simplification;234
8.7.6;5 Conclusion;234
8.7.7;Literature;235
8.8;Chapter 46. Distributed generation of synchronous, asynchronous, and
partially synchronous/asynchronous updating dynamics by a self-organizing oscillatory network;236
8.8.1;Abstract;236
8.8.2;1. INTRODUCTION;236
8.8.3;2. A SUMMARY OF THE SER ALGORITHM AND ITS PROPERTIES;236
8.8.4;3. THE SELF-ORGANIZING OSCILLATORY NETWORK (SOON) MODEL;237
8.8.5;4. ASYNCHRONOUS AND PARALLEL UPDATING;238
8.8.6;5. SYNCHRONOUS UPDATING;238
8.8.7;6. PAPS UPDATING;239
8.8.8;7. CONCLUSIONS;240
8.8.9;ACKNOWLEDGEMENTS;240
8.8.10;REFERENCES;240
8.9;Chapter 47. Direct Approaches to Improving the Robustness of Multilayer Neural
Networks;242
8.9.1;Abstract;242
8.9.2;1 Introduction;242
8.9.3;2 Definitions of robustness;242
8.9.4;3 Approach 1: Modification of the error function;243
8.9.5;4 Approach 2: Pruning and duplication;243
8.9.6;5 Discussion;245
8.9.7;6 Conclusion;245
8.9.8;7 References;245
8.10;Chapter 48. A Study on Generalization Properties of Artificial Neural Network Using Fahlman and Lebieres Learning Algorithm;246
8.10.1;Abstract;246
8.10.2;1. INTRODUCTION;246
8.10.3;2. DESCRIPTION OF THE ALGORITHM;246
8.10.4;3. SIMULATION RESULTS AND DISCUSSIONS;247
8.10.5;4. CONCLUSIONS;249
8.10.6;5. REFERENCES;249
8.11;Chapter 49. Domain Independent Testing and Performance Comparisons for Neural Networks;250
8.11.1;Abstract;250
8.11.2;1. INTRODUCTION;250
8.11.3;2. THE DATA GENERATING PROCESSES;251
8.11.4;3. PARAMETER SETTING FOR THE EXAMPLES;252
8.11.5;4. SIGNIFICANCE OF RESULTS;253
8.11.6;5. DOMAIN DEPENDENCE;254
8.11.7;6. REFERENCES;254
8.12;Chapter 50. The Optimal Elastic Net: Finding Solutions to the
Travelling Salesman Problem;256
8.12.1;Abstract;256
8.12.2;1 Introduction;256
8.12.3;2 The Optimal Elastic Net Method;258
8.12.4;3 Results;258
8.12.5;Conclusion;259
8.12.6;References;259
8.13;Chapter 51. A Bayesian Network for Temporal Segmentation;260
8.13.1;Abstract;260
8.13.2;1. INTRODUCTION;260
8.13.3;2. METHODS;260
8.13.4;3. EXAMPLE;262
8.13.5;4. RESULTS;262
8.13.6;References;263
8.14;Chapter 52. Adaptive constrained optimisation for improving the topological
maps;264
8.14.1;Abstract;260
8.14.2;1. INTRODUCTION;260
8.14.3;2. METHODS;260
8.14.4;3. EXAMPLE;262
8.14.5;4. RESULTS;262
8.15;Chapter 53. Adaptive constrained optimisation for improving the topological maps;264
8.15.1;Abstract;264
8.15.2;1. INTRODUCTION;264
8.15.3;2. ALGORITHMS;264
8.15.4;3. EXPERIMENTS;266
8.15.5;4. DISCUSSION;267
8.15.6;5. REFERENCES;267
9;Part V: Signal Processing;268
9.1;Chapter 54. A Recurrent Neural Network Model;270
9.1.1;Abstract;270
9.1.2;1. Partially recurrent networks;270
9.1.3;2. An illustration example;271
9.1.4;3. Concluding remarks;272
9.1.5;4. References;273
9.2;Chapter 55. The Minimum Entropy Neuron- a new building block for clustering transformations;274
9.2.1;Abstract;274
9.2.2;1. INTRODUCTION;274
9.2.3;2. THE MINIMUM ENTROPY PRINCIPLE;275
9.2.4;3. THE MINIMUM ENTROPY NEURON;275
9.2.5;4. EXAMPLE;276
9.2.6;5. DISCUSSION AND CONCLUSION;277
9.2.7;6. REFERENCES;277
9.3;Chapter 56. Nonlinear Hebbian Algorithms for Sinusoidal
Frequency Estimation;278
9.3.1;Abstract;278
9.3.2;1 Introduction;278
9.3.3;2 Nonlinear algorithms;279
9.3.4;3 Experimental results;279
9.3.5;References;281
9.4;Chapter 57. Systolic implementation of the orthogonal-inverse updating algorithm;282
9.4.1;Abstract;282
9.4.2;1. INTRODUCTION;282
9.4.3;2. ORTHOGONAL-INVERSE UPDATING ALGORITHM [9];283
9.4.4;3. SYSTOLIC IMPLEMENTATION;283
9.4.5;4. REFERENCES;285
9.5;Chapter 58. The Neural Impulse Response Filter;286
9.5.1;Abstract;286
9.5.2;1 Introduction;286
9.5.3;2 The NIR Filter;287
9.5.4;3 Magnetoencephalography;289
9.5.5;4 Conclusion;291
9.5.6;Appendix;291
9.5.7;References;291
9.6;Chapter 59. Stabilization Properties of Multilayer Feedforward
Networks with Time-Delay Synapses;292
9.6.1;Abstract;292
9.6.2;1 Introduction;292
9.6.3;2 The IIR Multilayer Perceptron;292
9.6.4;3 Stability Properties of the IIR MLP;293
9.6.5;4 Stabilization Experiment;294
9.6.6;5 Conclusions;295
9.6.7;6 References;295
9.7;Chapter 60. A Measure of Nonlinearity in Time Series Using
Neural Network Prediction Model;296
9.7.1;Abstract;296
9.7.2;1. INTRODUCTION;296
9.7.3;2. DEFINITION OF THE MEASURE;296
9.7.4;3. STATISTICAL TEST OF NONLINEARITY;297
9.7.5;4. Examples of Measuring Nonlinearity;297
9.7.6;5. CONCLUSIONS;299
9.7.7;6. REFERENCES;299
10;Part VI: Pattern Recognition 1;300
10.1;Chapter 61. A neural network approach to fault location in nonlinear dc circuits;302
10.1.1;Abstract;302
10.1.2;1. INTRODUCTION AND BACKGROUND;302
10.1.3;2. SIMULATION RESULTS;303
10.1.4;3. CONCLUSIONS;305
10.1.5;4. REFERENCES;305
10.2;Chapter 62. Optimization of the Distance-Based Neural Network
by Akaike's Information Criterion;306
10.2.1;Abstract;306
10.2.2;1. INTRODUCTION;306
10.2.3;2. OPTIMIZATION OF THE DISTANCE-BASED NEURAL NETWORK;307
10.2.4;3. EXPERIMENTAL;309
10.2.5;4. APPLICATION TO FACE RECOGNITION;311
10.2.6;5. CONCLUSION;312
10.2.7;Acknowledgements;312
10.2.8;References;312
10.3;Chapter 63. Loss Function Based Neural Classifiers;314
10.3.1;Abstract;314
10.3.2;1 Bayesian formulation of classification problem;314
10.3.3;2 Loss function approximations;315
10.3.4;3 Relationship to other work;316
10.3.5;4 Computational experiments;317
10.3.6;References;317
10.4;Chapter 64. Image Compression for Neural Networks using Chebyshev Polynomials;318
10.4.1;Abstract;318
10.4.2;1. INTRODUCTION;318
10.4.3;2. CHEBYSHEV APPROXIMATION;318
10.4.4;3. EQUALISING VARIANCES;319
10.4.5;4. HIGHER DIMENSIONS;320
10.4.6;5. EVEN AND ODD COEFFICIENTS;321
10.4.7;6. WEIGHTED COMPRESSION;321
10.4.8;7. CONCLUSION;321
10.4.9;REFERENCES;321
10.5;Chapater 65. Training the Gradient Field of a Dynamic Hopfield (Recurrent)
Network for Classification;322
10.5.1;1 Introduction;322
10.5.2;2 Algorithm for Design of Basins of Attraction;323
10.5.3;3 Simulation;324
10.5.4;4 Conclusion;324
10.5.5;References;325
10.6;Chapter 66. Classification of Epileptic EEG by Using Self-Organizing Maps;326
10.6.1;Abstract;326
10.6.2;1. Introduction;326
10.6.3;2. EEG Analysis Methods;327
10.6.4;3. Feature Classification;327
10.6.5;4. Results and Discussion;328
10.6.6;5. References;329
10.7;Chapter 67. Protein Structure Prediction and Neural Networks;330
10.7.1;1. Introduction;330
10.7.2;2. The state of the art;330
10.7.3;3. Current work;331
10.7.4;4. Limitations;331
10.7.5;5. Results;331
10.7.6;6. Conclusions;332
10.7.7;7. References;333
10.8;Chapter 68. A Neural Network for Meteor Trail Classification;334
10.8.1;Abstract;334
10.8.2;1. INTRODUCTION;334
10.8.3;2. PREVIOUS METHODS OF METEOR TRAIL CLASSIFICATION;335
10.8.4;3. DATA SOURCE;335
10.8.5;4. NEURAL NETWORK IMPLEMENTATION AND TRAINING;336
10.8.6;5. NETWORK VERIFICATION;336
10.8.7;6. CONCLUSION;337
10.8.8;7. REFERENCES;337
10.9;Chapter 69. High-speed Learning in a Supervised, Self-growing
Net;338
10.9.1;Abstract;338
10.9.2;1. THE NET;338
10.9.3;2. RESULTS;341
10.9.4;3. REFERENCES;341
10.10;CHAPTER 70. PARALLEL PROCESSING SYSTEM WITH FIXED CONNECTIONS AS A NEW
APPROACH TO HANDWRITTEN DIGIT RECOGNITION;342
10.10.1;Abstract;342
10.10.2;1. PREPROCESSING PART OF RECOGNITION SYSTEM;342
10.10.3;2. IDENTIFICATION AS A RECOGNITION PROCEDURE;344
10.10.4;3. PRELIMINARY RESULTS AND CONCLUSION;345
10.10.5;4. REFERENCES;345
10.11;CHAPTER 71. A MODIFIED HYPERMAP ARCHITECTURE FOR CLASSIFICATION OF BIOLOGICAL SIGNALS;346
10.11.1;Abstract;346
10.11.2;1. Introduction;346
10.11.3;2. The modified Hypermap Architecture;346
10.11.4;3. The learning algorithm;347
10.11.5;4. Results;348
10.11.6;Acknowledgement;348
10.11.7;References;349
11;Part VII: Applications and Pattern Recognition 2;350
11.1;Chapter 72. Texture image classification
based on interframe principal features;352
11.1.1;Abstract;352
11.1.2;1. INTRODUCTION;352
11.1.3;2. STRUCTURE OF THE TEXTURE IMAGE CLASSIFIER;353
11.1.4;3. TRAINING SYSTEM FOR THE CLASSIFICATION NETWORK;354
11.1.5;4. EXPERIMENTAL RESULTS;354
11.1.6;5. Conclusion;355
11.1.7;References;355
11.2;Chapter 73. Interframe principal feature extraction
using a multi-layer feedforward neural network;356
11.2.1;Abstract;356
11.2.2;1. INTRODUCTION;356
11.2.3;2. PRINCIPAL COMPONENT TRANSFORMATION;357
11.2.4;4. EXPERIMENTAL RESULTS;358
11.2.5;5. CONCLUSION;359
11.2.6;REFERENCES;359
11.3;Chapter 74. A method for analyzing decision regions in Learning Vector Quantization algorithms;360
11.3.1;Abstract;360
11.3.2;1. INTRODUCTION;360
11.3.3;2. TYPES OF CODEBOOK VECTORS;361
11.3.4;3. ANALYSIS OF MULTI-DIMENSIONAL DATA;362
11.3.5;4. CONCLUSIONS;363
11.3.6;5. REFERENCES;363
11.4;CHAPTER 75. TRACKING PARTICLES IN A HIGH ENERGY PARTICLE DETECTOR USING A
NEURAL GAS NETWORK;364
11.4.1;Abstract;364
11.4.2;1. INTRODUCTION;364
11.4.3;2. THE NETWORK;365
11.4.4;3. RESULT OF A SIMULATION.;366
11.4.5;4. REFERENCES;366
11.5;Chapter 76. Hand-written Japanese Kanji character recognition
by a structured self-growing neural network "CombNET-II";368
11.5.1;ABSTRACT;368
11.5.2;1. INTRODUCTION;368
11.5.3;2. A SELF GROWING NEURAL NETWORK MODEL "CombNET-II";369
11.5.4;3. HAND-WRITTEN KANJI CHARACTER RECOGNITION;370
11.5.5;4. CONCLUSIONS;370
11.5.6;ACKNOWLEDGMENTS;371
11.5.7;REFERENCES;371
11.6;Chapter 77. Development of MLP/LVQ hybrid networks for classification of
remotely-sensed satellite images;372
11.6.1;Abstract;372
11.6.2;1. INTRODUCTION;372
11.6.3;2. METHODOLOGY;373
11.6.4;3. RESULTS;374
11.6.5;4. CONCLUSIONS;375
11.6.6;5. REFERENCES;375
11.7;Chapter 78. A Deformable Templates Approach for Track
Finding;376
11.7.1;Abstract;376
11.7.2;1 Introduction;376
11.7.3;2 The Deformable Templates Method;376
11.7.4;3 Track Finding in High Energy Physics;378
11.7.5;4 Improved Gradient Descent;379
11.7.6;5 Summary;379
11.7.7;References;379
11.8;Chapter 79. Pattern Recognition with Artificial Neural Networks
- a Benchmark Study of Scaling Behaviour;380
11.8.1;Abstract;380
11.8.2;1 Introduction;380
11.8.3;2 The Problem;380
11.8.4;3 Networks;381
11.8.5;4 Theoretical Considerations;381
11.8.6;5 Simulations;382
11.8.7;6 Results;382
11.8.8;7 Conclusions and Discussion;384
11.8.9;References;384
11.9;Chapter 80. Reconstruction of Tokamak Density Profiles using Feedforward Networks;386
11.9.1;Abstract;386
11.9.2;1 Introduction;386
11.9.3;2 Neural Network Approach;387
11.9.4;3 Results and Discussion;389
11.9.5;References;389
11.10;CHAPTER 81. PATTERN RECOGNITION WITH THE RANDOM NEURAL NETWORK;390
11.10.1;Abstract;390
11.10.2;1 Introduction;390
11.10.3;2 The Random Neural Network as an auto-associative memory;391
11.10.4;3 The results;392
11.10.5;4 References;393
11.11;Chapter 82. Designing Modular Network Architectures Using a Genetic Algorithm;394
11.11.1;Abstract;394
11.11.2;1. INTRODUCTION;394
11.11.3;2. NETWORK DESIGN PRINCIPLES;395
11.11.4;3. SIMULATION AND RESULTS;396
11.11.5;4. REFERENCES;397
11.12;Chapter 83. On Clustering Properties of Hierarchical Self-Organizing Maps;398
11.12.1;Abstract;398
11.12.2;1 Introduction;398
11.12.3;2 Multilayer self-organizing map;398
11.12.4;3 Experimental Results;399
11.12.5;References;401
11.13;Chapter 84. SOC: A Self-Organizing Classifier;402
11.13.1;Abstract;402
11.13.2;1. INTRODUCTION;402
11.13.3;2. CLASSIFICATION AND SELF-ORGANIZATION;403
11.13.4;3. REFERENCES;405
11.14;Chapter 85. Cumulant-Based Neural Network Classifiers;406
11.14.1;Abstract;406
11.14.2;1. INTRODUCTION;406
11.14.3;2. A NEW INVARIANT REPRESENTATION;407
11.14.4;3. STRUCTURED NEURAL NETWORKS FOR IMAGE CLASSIFICATION;408
11.14.5;4. SIMULATION RESULTS;408
11.14.6;5. REFERENCES;409
11.15;Chapter 86. Model based approach for generating and structuring a learning database- for real-scale 3D identification tasks;410
11.15.1;Abstract;410
11.15.2;1- The real scale problem.;410
11.15.3;2- Model-based approach to generate a learning database;410
11.15.4;3- Formatting the database.;411
11.15.5;4- Experimental results;412
11.15.6;5- Conclusion;413
11.15.7;6- References;413
12;Part VIII: Software;414
12.1;Chapter 87. Neural Network Programming Environments;416
12.1.1;Abstract;416
12.1.2;1. Introduction;416
12.1.3;2. Application-oriented;418
12.1.4;3. Algorithm-oriented;419
12.1.5;4. General Purpose Programming Systems;420
12.1.6;5. Summary;423
12.1.7;6. References;423
12.2;Chapter 88. A Structured Design, Development and Integration Methodology
for Real-World Applications of Artificial Neural Networks;424
12.2.1;Abstract;424
12.2.2;1 INTRODUCTION;424
12.2.3;2 ASSESSMENT PHASE;425
12.2.4;3 SPECIFICATION PHASE: THE APPLICATION DOMAIN;426
12.2.5;4 DESIGN PHASE: ENGINEERING THE SYSTEM;427
12.2.6;5 IMPLEMENTATION PHASE: CONSTRUCTING THE SYSTEM;428
12.2.7;6 EVALUATION PHASE: TESTING THE PERFORMANCE OF THE SYSTEM;429
12.2.8;7 DELIVERY PHASE: INSTALLING THE SYSTEM;429
12.2.9;8 CONCLUSION;430
12.2.10;9 REFERENCES;430
12.3;Chapter 89. A new artificial neural network classifier;432
12.3.1;Abstract;432
12.3.2;1 INTRODUCTION;432
12.3.3;2 THE MLNSS NETWORK;433
12.3.4;3 THE MLNSup NETWORK;434
12.3.5;4 RESULTS;435
12.3.6;5 CONCLUSIONS;435
12.3.7;6 ACKNOWLEDGEMENTS;435
12.3.8;7 REFERENCES;435
12.4;Chapter 90. Software Package for Multilayer Perceptron
Simulation as Associative Memory;436
12.4.1;1 Abstract;436
12.4.2;2 Theoretical Background;436
12.4.3;3 Computer simulation results;438
12.4.4;4 Conclusions;438
12.4.5;References;439
12.5;Chapter 91. Efficient Simulation of Massive Neural Networks on Machines with Limited Memory;440
12.5.1;Abstract;440
12.5.2;1. INTRODUCTION;440
12.5.3;2. ACCESS INTO MULTIDIMENSIONAL LAYERS;441
12.5.4;3. A LOOKUP ALGORITHM FOR REGENERATING CONNECTIONS;441
12.5.5;4. PERFORMANCE;442
12.5.6;5. CONCLUSION;443
12.5.7;ACKNOWLEDGEMENTS;443
12.5.8;References;443
12.6;Chapter 92. SESAME — A software environment for combining multiple neural
network paradigms and applications;444
12.6.1;Abstract;444
12.6.2;Introduction;444
12.6.3;The Concept;444
12.6.4;Class hierarchy;445
12.6.5;SESAME's Network Description Language and Experiment Control;445
12.6.6;Examples: Backpropagation, Feature Maps, and others;445
12.6.7;Conclusion;446
12.6.8;References;446
12.7;Chapter 93.
A Design Framework for Neural Nets;448
12.7.1;Abstract;448
12.7.2;1. Introduction;448
12.7.3;2. Design System Overview;448
12.7.4;3. An Example;449
12.7.5;4. Applications of the Design System;451
12.7.6;References;451
12.8;Chapter 94. Using a Library of Efficient Data Structures and Algorithms
as a Neural Network Research Tool;452
12.8.1;Abstract;452
12.8.2;1 Introduction;452
12.8.3;2 Basic Features of LEDA;453
12.8.4;3 LEDA and Neural Networks;454
12.8.5;4 Realizing the Growing Cell Structures with LEDA;454
12.9;Chapter 95. Transforming neural network specifications to
parallel programs;456
12.9.1;Abstract;456
12.9.2;1. INTRODUCTION;456
12.9.3;2. TRANSFORMATIONS TO OBTAIN A SEQUENTIAL PROGRAM;457
12.9.4;3. TRANSFORMATIONS TO OBTAIN PARALLEL PROGRAMS;459
12.9.5;4. DISCUSSION AND FUTURE WORK;459
12.9.6;5. REFERENCES;460
12.9.7;6 . EMPLOYED BASIC LAWS;460
12.9.8;7. ACKNOWLEDGEMENTS;460
12.10;Chapter
96. The Implementation of Artificial Neural Networks On Parallel Heterogeneous Architectures;462
12.10.1;Abstract;462
12.10.2;Parallel Computer Systems;462
12.10.3;N-Tuple Network Implementation;462
12.10.4;Analogue Neural Networks and Genetic Algorithms;464
12.10.5;Parallel Heterogeneous Network Implementation;465
12.10.6;Conclusions;465
12.10.7;Acknowledgements.;465
12.10.8;References;465
12.11;Chapter 97. Implementations of Very Large Recurrent ANNs
on Massively Parallel SIMD Computers;466
12.11.1;Abstract;466
12.11.2;Introduction;466
12.11.3;Remarks on Scaling and Computational Complexity;467
12.11.4;Implementations;468
12.11.5;Discussion;469
12.11.6;References;469
12.12;Chapter
98. An Acoustical Signal Recognizer Implemented on a Novel Interactive Object-Oriented Neural Network Simulator;470
12.12.1;Abstract;470
12.12.2;1. The Recognition of Acoustical Signals;470
12.12.3;2. The Simulator;471
12.12.4;3. Achieved Results and Conclusions;473
12.12.5;4. References;473
13;Part IX: Cognitive Systems;474
13.1;Chapter
99. A Connectionist Approach to Solving Tangram Puzzles;476
13.1.1;1 Introduction;476
13.1.2;2 A Connectionist Model for Grid Tangrams;477
13.1.3;References;479
13.2;Chapter
100. A Connectionist Model of Pure Alexia;480
13.2.1;Abstract;480
13.2.2;1. PURE ALEXIA;480
13.2.3;2. THE MODEL;481
13.2.4;3. THE SIMULATIONS;482
13.2.5;4. DISCUSSION;483
13.2.6;5. REFERENCES;483
13.3;Chapter
101. Cognition and Neural Network Modelling;484
13.3.1;Abstract;484
13.3.2;1. Introduction;484
13.3.3;2. Data representation;484
13.3.4;3. Cognition;485
13.3.5;4. Neural Network Model;485
13.3.6;5. Conclusion;488
13.3.7;References;488
13.4;Chapter
102. Counting With Artificial Neural Networks: An Experiment;490
13.4.1;1 Introduction;490
13.4.2;2 Mixed architectures with delayed links;491
13.4.3;3 Counting to 16 items;492
13.4.4;References;492
13.5;Chapter
103. Combining Supervised and Unsupervised Learning in a Model of Categorisation;494
13.5.1;Abstract;494
13.5.2;1. INTRODUCTION;494
13.5.3;2) THE DOMAINS ACCOUNT - UNSUPERVISED LEARNING;495
13.5.4;3) EXTENSION TO SUPERVISED LEARNING;495
13.5.5;4) COMBINING SUPERVISED AND UNSUPERVISED LEARNING;496
13.5.6;5) CONCLUSION;497
13.5.7;6) REFERENCES;498
13.6;Chapter
104. On the beauty of nature;500
13.6.1;Abstract;500
13.6.2;1. INTRODUCTION;500
13.6.3;2. THE
MODEL;501
13.6.4;3. SIMULATION RESULTS;502
13.6.5;4. DISCUSSION;503
13.6.6;5. REFERENCES;503
13.7;Chapter
105. Basins of Attraction in Disordered Networks;504
13.7.1;Abstract;504
13.7.2;1. Introduction;504
13.7.3;2. Disordered CA architecture;504
13.7.4;3. Basins of Attraction;505
13.7.5;4. Computing Pre — images;506
13.7.6;5. Brain—Like Computation;508
13.7.7;6. A Mind Model;508
13.7.8;7. Learning;508
13.7.9;References;509
13.8;Chapter
106. Sequential learning in mean field autoassoeiators;510
13.8.1;Abstract;510
13.8.2;1 INTRODUCTION;510
13.8.3;2 SEQUENTIAL LEARNING;510
13.8.4;3 MEAN FIELD AUTOASSOCIATORS;511
13.8.5;4 CONCLUSIONS;511
13.8.6;5 REFERENCES;512
13.9;CHAPTER
107. VARIABILITY IN THE PATHOGENESIS OF ALZHEIMER DISEASE: ANALYTICAL RESULTS;514
13.9.1;Abstract;514
13.9.2;1. INTRODUCTION;514
13.9.3;2. A MODEL OF DELETION AND COMPENSATION;514
13.9.4;3. SOLUTION OF THE MODEL;515
13.9.5;4. DISCUSSION;516
13.9.6;REFERENCES;517
13.10;Chapter
108. Microeconomic experiments by neural networks;518
13.10.1;Abstract;518
13.10.2;1. INTRODUCTION;518
13.10.3;2. CROSS-TARGETS;519
13.10.4;3. FROM A SIMPLE
SUBJECT;520
13.10.5;4. TO MORE COMPLEX SUBJECTS;520
13.10.6;5. FURTHER DEVELOPMENTS;521
13.10.7;6. REFERENCES;521
13.11;Chapter
109. Interacting neural networks: an artificial life approach for stock markets;522
13.11.1;Abstract;522
13.11.2;1. INTRODUCTION;522
13.11.3;2. THE GENERAL FRAMEWORK;523
13.11.4;3. THE LEARNING PARADIGMS;523
13.11.5;4. CONCLUSION;525
13.11.6;5. REFERENCES;525
13.12;Chapter
110. Resolving Linguistic Ambiguities with a Neural Data-Oriented Parsing (DOP) System;526
13.12.1;Abstract;526
13.12.2;1. Background;526
13.12.3;2. Introduction;527
13.12.4;3. The Model;528
13.12.5;4. Results;528
13.12.6;5. Discussion and Conclusions;528
13.12.7;6. References;529
13.13;Chapter
111. Implementing a Semantic Subsystem for a Model of Language Acquisition;530
13.13.1;Abstract;530
13.13.2;1. INTRODUCTION;530
13.13.3;2. A MODEL OF LANGUAGE ACQUISITION;530
13.13.4;3. SEMANTIC SUBSYSTEM;531
13.13.5;4. CONCLUSIONS;533
13.13.6;5. FURTHER WORK;534
13.13.7;6. ACKNOWLEDGEMENTS;534
13.13.8;7. REFERENCES;534
13.14;Chapter
112. Finding Multi-Faculty Structure;536
13.14.1;Abstract;536
13.14.2;1. INTRODUCTION;536
13.14.3;2. THE ARCHITECTURE;536
13.14.4;3. OPERATION OF THE MODEL;537
13.14.5;4. DISCUSSION AND CONCLUSIONS;538
13.14.6;5. APPENDIX A - Representing continuous values;539
13.14.7;6. APPENDIX B - Experimental Parameters;539
13.14.8;6. REFERENCES;539
13.15;Chapter 113. A Connectionist Approach to Anaphora Resolution
in Task-Oriented Discourses;540
13.15.1;Abstract;540
13.15.2;1. RESEARCH AIM;540
13.15.3;2. OUTLINING THE SOLUTION;542
13.15.4;References;543
13.16;Chapter
114. Unsupervised methods for finding linguistic categories;544
13.16.1;Abstract;544
13.16.2;1. INTRODUCTION;544
13.16.3;2. NETWORK SIMULATIONS;546
13.16.4;3. BENCHMARKING NETWORK PERFORMANCE;547
13.16.5;4. CONCLUSION;547
13.16.6;5. REFERENCES;547
13.17;Chapter
115. The Speed and Slips of Mental Arithmetic;548
13.17.1;Abstract;548
13.17.2;Introduction;548
13.17.3;Architecture;548
13.17.4;Simulations;549
13.17.5;Results;549
13.17.6;Analysis and discussion;550
13.17.7;Acknowledgements;551
13.17.8;References;551
13.18;Chapter
116. Looks recognition by Adaptive Junction;552
13.18.1;Abstract;552
13.18.2;1 Introduction;552
13.18.3;2 Adaptive Junction;552
13.18.4;3 Simulation;553
13.18.5;4 Discussion;553
13.18.6;5 Conclusion;555
13.18.7;References;555
13.19;Chapter
117. A Connectionist Approach to Effects of Anxiety and Task Difficulty on Learning;556
13.19.1;Abstract;556
13.19.2;1. INTRODUCTION;556
13.19.3;2. DESCRIPTION AND REQUIREMENTS OF AN APPROPRIATE FRAME
WORK;556
13.19.4;3. COMPUTATIONAL MODEL;557
13.19.5;4. RESULTS AND DISCUSSION;558
13.19.6;Acknowledgements;558
13.19.7;REFERENCES;559
13.20;Chapter 118. A Neural Network Architecture which uses Cellular Automata to get Context-Specific Associative
Memory;560
13.20.1;Abstract;560
13.20.2;1. Introduction;560
13.20.3;2. The used neural network model with Cellular Automata
behaviour;561
13.20.4;3. Building up a context-specific associative memory;562
13.20.5;4. Simulation results;563
13.20.6;References;563
13.21;Chapter
119. Temporal order, timing, and probability context effects on pattern recognition and categorization in neural networks;564
13.21.1;Abstract;564
13.21.2;1 INTRODUCTION;564
13.21.3;2 MODEL DESCRIPTION;565
13.21.4;3 Network Equations;567
13.21.5;4 Simulation Results;568
13.21.6;Acknowledgements;568
13.21.7;References;568
13.22;Chapter
120. Distributed Representation: A critique of minimal networks;570
13.22.1;Abstract;570
13.22.2;1. INTRODUCTION;570
13.22.3;2. TWO THEORETICAL FRAMEWORKS;570
13.22.4;3. DISTRIBUTED REPRESENTATION;571
13.22.5;4. MINIMAL NETWORKS;572
13.22.6;5. PRACTICAL IMPLICATIONS;573
13.22.7;6. NOTES;573
13.22.8;7. REFERENCES;573
14;Part XI: Hardware;574
14.1;Chapter 121. To simulate or not to
simulate;576
14.1.1;Abstract;576
14.1.2;1. INTRODUCTION;576
14.1.3;3. Implications for hardware systems;579
14.1.4;4. Comparison of performance of simulations;579
14.1.5;Character recognition system;580
14.1.6;6. Conclusions;581
14.1.7;7. Acknowledgements;582
14.1.8;8. References;582
14.2;Chapater
122. Three-Dimensional Neural Network Synthesis;584
14.2.1;Abstract;584
14.2.2;1. THE PROBLEM;584
14.2.3;2. THE NETWORK;585
14.2.4;3. THE NEURON;586
14.2.5;4. REFERENCES;587
14.3;Chapter
123. Hopfield's Binary Networks: Simulations on Tree-Connected Transputer Networks;588
14.3.1;Abstract;588
14.3.2;1. INTRODUCTION;588
14.3.3;2. EXPLOITING THE CONNECTION PARALLELISM;589
14.3.4;3. THE ON-TREE SOLUTION: EXPERIMENTAL FINDINGS;590
14.3.5;4. CONCLUDING REMARKS;592
14.3.6;REFERENCES;592
14.4;Chapter
124. Further Developments of a Ring Based Multiprocessor Architecture for Back Prop Simulation;594
14.4.1;Abstract;594
14.4.2;1. INTRODUCTION;594
14.4.3;2. HOST TIMING CONSTRAINTS;595
14.4.4;3. A SIMPLIFIED INSTRUCTION SET;596
14.4.5;4. REFERENCES;597
14.5;Chapter
125. The NAND-Net — A Binary Connectionist Model;598
14.5.1;Abstract;598
14.5.2;1 Introduction;598
14.5.3;2 The Architecture of NAND-Nets;598
14.5.4;3 NAND-Nets and Learning;599
14.5.5;4 Discussion of NAND-Nets;600
14.5.6;5 Conclusions;601
14.5.7;References;601
14.6;Chapater 126. Adaptable VLSI Neural Network of Tens of
Thousand Connections;602
14.6.1;ABSTRACT;602
14.6.2;1. INTRODUCTION;602
14.6.3;2. ANALOGUE-DIGITAL HYBRID VLSI NEURAL
NETWORK;603
14.6.4;3.
CONCLUSION /DISCUSSION;605
14.6.5;4. REFERENCES;605
14.7;Chapter 127.
Hardware Implementations of PC A Neural Networks;606
14.7.1;Abstract;606
14.7.2;1. PCA Neural Networks;606
14.7.3;2. The Weighted Subspace Network;606
14.7.4;3. Implementation with a Digital Signal Processor;607
14.7.5;4. Special-Purpose Bit-Serial Hardware with Parallel Extension Capability;607
14.7.6;5. References;609
14.8;CHAPTER 128. VLSI ARCHITECTURE OF THE SELF-ORGANIZING NEURAL NETWORK USING SYNCHRONOUS
PULSE-DENSITY MODULATION TECHNIQUE;610
14.8.1;Abstract;610
14.8.2;1. INTRODUCTION;610
14.8.3;2. ARCHITECTURE;610
14.8.4;3. CONCLUSIONS;613
14.8.5;REFERENCES;613
14.9;CHAPTER 129. A PROCESSOR RING FOR THE IMPLEMENTATION OF NEURAL
NETWORKS WITH CORTICAL ARCHITECTURE;614
14.9.1;Abstract;614
14.9.2;1. Introduction;614
14.9.3;2. Description of the Processor Ring;615
14.9.4;3. Performance and Discussion;616
14.9.5;4. References;617
14.10;Chapter 130
. The Dynamic Ring Architecture;618
14.10.1;1. INTRODUCTION;618
14.10.2;2. THE DYNAMIC RING
ARCHITECTURE;618
14.10.3;3. NEURAL NETWORK EMULATION;619
14.10.4;4. THE UTAK1 PROCESSOR;621
14.10.5;5. REFERENCES;621
14.11;CHAPTER 131. PARALLEL HARDWARE IMPLEMENTATION OF THE
GENERAL NEURAL UNIT MODEL;622
14.11.1;Abstract;622
14.11.2;1. BACKGROUND AND MOTIVATION;622
14.11.3;2. PARALLEL IMPLEMENTATION;622
14.11.4;3. DISCUSSION;624
14.11.5;4. REFERENCES;625
14.12;Chapter
132. Using Threshold Gates To Implement Sigmoid Nonlinearity;626
14.12.1;Abstract;626
14.12.2;1. INTRODUCTION;626
14.12.3;2. THEORETICAL ASPECTS;626
14.12.4;3. OPTIMAL IMPLEMENTATION;627
14.12.5;4. APPROXIMATE MULTIPLICATION;628
14.12.6;5. CONCLUSIONS;629
14.12.7;6. REFERENCES;629
14.13;Chapter
133. Hardware Implementations of Kohonen's Selforganizing Feature Map;630
14.13.1;Abstract;630
14.13.2;1 Introduction;630
14.13.3;2 The implementation concepts of the neural coprocessor;631
14.13.4;3 The memory arithmetic board (MAB);632
14.13.5;4 System integration of the neural coprocessor;632
14.13.6;5 Results;632
14.13.7;6 Summary and Conclusions;633
14.13.8;References;633
14.14;Chapter 134. RENNS - a REconfigurable Neural Network
Server;634
14.14.1;Abstract;634
14.14.2;Introduction and motivation;634
14.14.3;Description of RENNS;634
14.14.4;Levels of reconfiguration;635
14.14.5;Possible configurations;636
14.14.6;Neural network application development on RENNS;636
14.14.7;Status and future work;638
14.14.8;References;638
14.15;Chapter
135. NAC: A Building Block for Neural Architectures;640
14.15.1;Abstract;640
14.15.2;1. BUILDING BLOCK;640
14.15.3;2. NAC APPLICATIONS;642
14.15.4;3. CONCLUSION;643
14.15.5;4. REFERENCES;643
14.16;Chapter
136. A Massively Parallel Implementation of Autoassociative Neural Networks for Image Compression;644
14.16.1;Abstract;644
14.16.2;1 Introduction;644
14.16.3;2 Neural Autoassociation for Image Compression;644
14.16.4;3 The Associative String Processor;645
14.16.5;4 Parallel Implementation;646
14.16.6;5 Conclusions;647
14.16.7;References;647
14.17;Chapter
137. A Faithful Analog Implementation of The Boltzmann Machine;648
14.17.1;Abstract;648
14.17.2;1. ALGORITHM;648
14.17.3;2. FUNCTIONAL CELLS;648
14.17.4;3 . RANDOM GENERATOR;650
14.17.5;4. EXPERIMENTAL RESULTS;650
14.17.6;5. CONCLUSION;650
14.17.7;6. REFERENCES;651
15;Part XII: Commercial / Industrial Hardware Systems;652
15.1;Chapter
138. A temporal noisy-leaky integrator neuron constructed using pRAMs;654
15.1.1;Abstract;654
15.1.2;1. INTRODUCTION;654
15.1.3;2. THE TEMPORALITY OF THE pRAM TNLI MODEL;654
15.1.4;3. DIGITAL HARDWARE IMPLEMENTATION;655
15.1.5;4. ASPECTS OF TRAINING AND LEARNING;657
15.1.6;6. CONCLUSION;657
15.1.7;7. REFERENCES;657
15.2;Chapter 139. Optical high order feedback neural network
using an optical fibre amplifier;658
15.2.1;Abstract;658
15.2.2;1. INTRODUCTION;658
15.2.3;2. THE HIGH ORDER FEEDBACK NEURAL NET (HOFNET);659
15.2.4;3. OPTICAL HOFNET SYSTEM;659
15.2.5;4. SIMULATIONS AND CONCLUSIONS;661
15.2.6;5. REFERENCES;661
15.3;Chapter
140. MindShape1: a neurocomputer concept based on a fractal architecture;662
15.3.1;Abstract;662
15.3.2;1. INTRODUCTION;662
15.3.3;2. THE FRACTAL ARCHITECTURE;663
15.3.4;3. EFFICIENCY OF FRACTAL ARCHITECTURES;663
15.3.5;4. IMPLEMENTATION PRINCIPLES;664
15.3.6;5. CONCLUDING REMARKS;665
15.3.7;6. REFERENCES;665
15.4;Chapter
141. Parallel Distributed Computation;666
15.4.1;Abstract;666
15.4.2;1 Introduction;666
15.4.3;2 Background;666
15.4.4;3 PDC within a state machine;667
15.4.5;4 Generalizing PDC;668
15.4.6;5 Conclusions;669
15.4.7;6 References;669
15.5;Chapter
142. A VLSI Implementation of Programmable Cellular Neural Networks;670
15.5.1;ABSTRACT;670
15.5.2;1. INTRODUCTION;670
15.5.3;2. THE ELEMENTARY NEURAL CELL;671
15.5.4;3. ARRAY ARCHITECTURE;671
15.5.5;References;672
15.6;Chapter
143. Radial Base Network for Dynamic Fault Detection;674
15.6.1;Abstract;674
15.6.2;1. INTRODUCTION;674
15.6.3;2. RADIAL BASE NETWORK;674
15.6.4;3. PARAMETER ESTIMATION;675
15.6.5;4. DYNAMIC FAULT DETECTION;676
15.6.6;5. SIMULATION EXAMPLE;676
15.6.7;6. CONCLUSIONS;677
15.6.8;7. REFERENCES;677
15.7;Chapter
144. A neural technique for satellite coverage plans optimization;678
15.7.1;Abstract;678
15.7.2;1. INTRODUCTION;678
15.7.3;2. THE CONSTRAINT SATISFACTION NETWORK;678
15.7.4;3. SATELLITE COVERAGE PLANS OPTIMIZATION;679
15.7.5;4. CONCLUSIONS;680
15.7.6;5. REFERENCE;680
15.8;Chapter
145. Integrated cooperation between unsupervised and supervised learning for a defect classification problem;682
15.8.1;Abstract;682
15.8.2;1. INTRODUCTION;682
15.8.3;2. CONNECTIONIST MODELS FOR VISION;682
15.8.4;3. AUTOMATIC DEFECT CLASSIFICATION FOR QUALITY CONTROL;683
15.8.5;4. DESCRIPTION OF THE MODEL;684
15.8.6;5. RESULTS;685
15.8.7;6. DISCUSSION;686
15.8.8;8. REFERENCES;686
15.9;CHAPTER
146. NEURAL NETWORK IMAGE SEGMENTATION FOR AUTOMATED VISUAL INSPECTION;688
15.9.1;Abstract;688
15.9.2;1. Texture Segmentation with Neural Networks;688
15.9.3;2. Experimental Setup and Results;689
15.9.4;Conclusion;691
15.9.5;References;691
15.10;Chatper 147. A Comparison Between Chemotaxis and Back-Propagation
Learning Applied to Colour Recipe Prediction.;692
15.10.1;Abstract;692
15.10.2;1. INTRODUCTION;692
15.10.3;2. BACK-PROPAGATION NETWORKS AND RECIPE PREDICTION;693
15.10.4;3. CHEMOTAXIS: A SIMPLER LEARNING ALGORITHM;694
15.10.5;4. CONCLUSION;695
15.10.6;5. REFERENCES;695
15.11;Chapter
148. Dynamic Model for A Plant Using Associative Memory System;696
15.11.1;Abstract;696
15.11.2;1. INTRODUCTION;696
15.11.3;2. THE SEWAGE PUMP STATION PLANT SYSTEM;696
15.11.4;3. PUMP OPERATION BY MEANS OF SFK;697
15.11.5;4. WATER LEVEL PATTERNS GENERATED BY MEANS OF DFK;698
15.11.6;5. CONCLUSION;699
15.11.7;REFERENCES;699
15.12;CHAPTER
149. LARGE SCALE APPLICATION OF NEURAL NETWORKS TO PROTEIN CLASSIFICATION;700
15.12.1;Abstract;700
15.12.2;1. Introduction;700
15.12.3;2. Large scale application;700
15.12.4;3. Hybridation of statistical and ANN approaches;702
15.12.5;References;703
15.13;CHAPTER
150. AN ORDERING THEOREM THAT ALLOWS FOR ORDERING CHANGES;704
15.13.1;Abstract;704
15.13.2;1. Introduction;704
15.13.3;2. An ordering theorem for synaptic changes of finite size.;706
15.13.4;3. Discussion;707
15.13.5;References;708
15.14;Chapter
151. Process State Monitoring Using Self-Organizing Maps;710
15.14.1;Abstract;710
15.14.2;1. INTRODUCTION;710
15.14.3;2. EXPERIMENTS;711
15.14.4;3. CONCLUSIONS;712
15.14.5;References;712
15.15;Chapater
152. The Hardware Implementation of Logical Neural Networks in an Ultrasonic Weld Control System;714
15.15.1;Abstract;714
15.15.2;1. INTRODUCTION AND EVALUATION OF THE EXISTING SAW CONTROLLER;714
15.15.3;2. DESIGN OF THE HARDWARE PREPROCESSOR;715
15.15.4;3. EXPERIMENTS WITH THE PROPOSED ARCHITECTURE;716
15.15.5;4. DISCUSSION;716
15.15.6;5. CONCLUSIONS.;717
15.15.7;6. REFERENCES;717
15.16;Chapter
153. The control of submerged arc welding using neural network interpretation of ultrasound;718
15.16.1;Abstract;718
15.16.2;1. INTRODUCTION;718
15.16.3;2. DATA COLLECTION METHOD;718
15.16.4;3. NEURAL NETWORK ARCHITECTURE;719
15.16.5;4. TRAINING AND OPTIMISATION;721
15.16.6;5. RESULTS;721
15.16.7;6. CONCLUSIONS;721
15.16.8;7. REFERENCES;722
16;Part XIII: Algorithms and Applications 1;724
16.1;Chapter 154.
Reliability in Neural Network Learning;726
16.1.1;Abstract;726
16.1.2;1 Introduction;726
16.1.3;2 Provably Reliable and Efficient Neural Classifiers;728
16.1.4;3 Neural Improvement of Approximations of Realvalued
Functions;731
16.1.5;4 Dynamical Selection of Topologies for Reliable Networks;734
16.1.6;5 Conclusion;736
16.1.7;6 Acknowledgements;736
16.1.8;References;736
16.2;CHAPTER 155. AUTOMATED RADAR BEHAVIOUR ANALYSIS USING HIERARCHICAL
NEURAL NETWORK ARCHITECTURES;738
16.2.1;Abstract;738
16.2.2;1. INTRODUCTION;738
16.2.3;2 PROJECT BACKGROUND;738
16.2.4;3. BACKGROUND: DATA FUSION;739
16.2.5;4. ASSESSMENT: ANN FOR DATA FUSION;739
16.2.6;5 SPECIFICATION: TRACKER'S FUNCTIONALITY;740
16.2.7;6. DESIGN AND IMPLEMENTATION: THE TRACKER DEMONSTRATOR;740
16.2.8;7 EVALUATION;742
16.2.9;8 CONCLUSION;742
16.2.10;REFERENCES;743
16.3;Chapter 156.
A Neural Network Approach to Restrictive Channel Routing Problems;744
16.3.1;Abstract;744
16.3.2;1. Introduction;744
16.3.3;2. Neural Network Approach;744
16.3.4;3. Simulation Results and Conclusion;746
16.3.5;4. References;747
16.4;Chapter
157. BPCLS: A Classified Learning Search Algorithm Based On Back-Propagation Model;748
16.4.1;Abstract;748
16.4.2;1. INTRODUCTION;748
16.4.3;2. THE BPCLS ALGORITHM;749
16.4.4;3. THE EFFICIENCY OF BPCLS;749
16.4.5;4. BACK-PROPAGATION MODEL IN BPCLS;750
16.4.6;5. CONCLUSION;751
16.4.7;6. REFERENCES;751
16.5;Chapter
158. Selection of Optimal Parameters for Kohonen Self-organizing Feature Maps;752
16.5.1;Abstract;752
16.5.2;1. One-dimensional Case;752
16.5.3;References;756
16.6;Chapter
159. Digit classification using an edge based hierarchical neural representation;758
16.6.1;Abstract;758
16.6.2;1. INTRODUCTION;758
16.6.3;2. RECORDING AND LOGARITHMIC TRANSFORMATION;759
16.6.4;3. CODING AND LINKING PROCESS;759
16.6.5;4. CLASSIFICATION;759
16.6.6;5. RESULTS;760
16.6.7;6. CONCLUSION;761
16.6.8;7. REFERENCES;761
16.7;Chapter
160. Selecting Reliable Kohonen Maps for Data Analysis;762
16.7.1;Abstract;762
16.7.2;KOHONEN MAPS AND DATA ANALYSIS;762
16.7.3;A SELECTION
PROCESS;763
16.7.4;CONCLUSION;765
16.7.5;REFERENCES;765
16.8;Chapter
161. A Logical Neural Network based Character Recognition System that Adapts to an Author's Writing Style;766
16.8.1;1. Introduction;766
16.8.2;2. Subclass Selection.;767
16.8.3;3. Results.;769
16.8.4;4. Conclusions.;769
16.8.5;5. References;769
16.9;Chapter 162. Optical Character Recognition and Cooperating Neural Networks
techniques;770
16.9.1;Abstract;770
16.9.2;1. Introduction;770
16.9.3;2. Description of the OCR process;770
16.9.4;3. Design of the data base;771
16.9.5;4. Description of the networks and experiments;771
16.9.6;5. Results;772
16.9.7;6. Conclusion;773
16.9.8;7. Bibliography;773
16.10;Chapter
163. Solving the Human Face Recognition Task using Neural Nets;774
16.10.1;Abstract;774
16.10.2;1. Introduction;774
16.10.3;2. Image feature extraction using MLP;774
16.10.4;3. Feature classification;775
16.10.5;4. Experimental results;776
16.10.6;5. Conclusion;777
16.10.7;Bibliography;777
16.11;Chapter
164. Scene segmentation using Multiresolution Analysis and MLP;778
16.11.1;Abstract;778
16.11.2;1. Introduction;778
16.11.3;2. Multiresolution Analysis;778
16.11.4;3. An identity recognition network for segmentation;779
16.11.5;4. A two class network;779
16.11.6;5. Conclusion;781
16.11.7;Bibliography;781
16.12;Chapter
165. Secondary structure of proteins from NMR data by neural nets.;782
16.12.1;Abstract;782
16.12.2;1. INTRODUCTION;782
16.12.3;2. Physical methods;783
16.12.4;3. The use of neural networks;783
16.12.5;4. Results;784
16.12.6;Acknowledgements;785
16.12.7;References;785
16.13;Chapter 166.
Comparative Study of Neural Networks and Non Parametric Statistical Methods for Off-Line Handwritten Character Recognition;786
16.13.1;Abstract;786
16.13.2;1. THE NETWORKS;786
16.13.3;2. EXPERIMENTAL RESULTS;788
16.13.4;3. DISCUSSION;788
16.13.5;ACKNOWLEDGEMENTS;789
16.13.6;REFERENCES;789
16.14;Chapter 167. Integration of a connectionist model in information retrieval
systems;790
16.14.1;Abstract;790
16.14.2;1. INTRODUCTION;790
16.14.3;2. AN ASSOCIATIVE MODEL FOR KNOWLEDGE REPRESENTATION;790
16.14.4;3. ASSOCIATIVE INFORMATION RESEARCH;792
16.14.5;4. THE DIFFERENT LEARNING METHODS;792
16.14.6;5. CONCLUSION;793
16.14.7;6. REFERENCES;793
16.15;Chapter
168. High Impedance Fault Detection using Recurrent Networks;794
16.15.1;Abstract;794
16.15.2;1. INTRODUCTION;794
16.15.3;2. DATA PRE-PROCESSING;795
16.15.4;3. TRAINING;796
16.15.5;4. RESULTS;796
16.15.6;5. CONCLUSION;797
16.15.7;6. REFERENCES;797
17;Part XIV: Algorithms and Applications 2;798
17.1;Chapter
169. Self-Organizing Feature Maps in Texture Classification and Segmentation;800
17.1.1;Abstract;800
17.1.2;1 Introduction;800
17.1.3;2 Self-Organizing Feature Map
in Pattern Recognition;801
17.1.4;3 Applications;802
17.1.5;References;806
17.2;Chapter 170. Prediction of monthly transition of the composition stock price index using recurrent
back-propagation;808
17.2.1;Abstract;808
17.2.2;1. INTRODUCTION;808
17.2.3;2. FORMATION OF THE PROBLEM;808
17.2.4;3. THE NETWORKS;809
17.2.5;4. EXPERIMENTS;810
17.2.6;5. CONCLUSION;810
17.2.7;6. REFERENCES;811
17.3;Chapter
171. EuclidNet - A Multilayer Neural Network using the Euclidian Distance as Propagation Rule;812
17.3.1;Abstract;812
17.3.2;1. THE PROBLEM - INTERMEDIATE PROTOTYPES;812
17.3.3;2. THE SOLUTION - EUCLIDNET;813
17.3.4;3. REFERENCES;815
17.4;CHAPTER
172. NEURAL NETWORK APPROACH FOR DETECTING HEART DISEASE;816
17.4.1;Abstract;816
17.4.2;1. INTRODUCTION;816
17.4.3;2. METHODS AND RESULTS;816
17.4.4;3. DISCUSSION;819
17.4.5;4. REFERENCES;819
17.5;CHAPTER
173. PATTERN CLASSIFICATION USING NEURAL NETWORKS FOR DIAGNOSTIC PROBLEM SOLVING;820
17.5.1;Abstract;820
17.5.2;1. INTRODUCTION;820
17.5.3;2. PATTERN RECOGNITION APPROACH;821
17.5.4;3. NEURAL NETWORK APPROACH;821
17.5.5;4. NEURAL NETWORK PATTERN CLASSIFICATION;821
17.5.6;5. CONCLUSIONS;822
17.5.7;6. REFERENCES;823
17.6;Chapter
174. Machine-generated music with themes;824
17.6.1;Abstract;824
17.6.2;1 INTRODUCTION;824
17.6.3;2 REPRESENTATION OF THE RULES;824
17.6.4;3 GENERALIZING THE RULES;825
17.6.5;4 CHOOSING WHAT RULE TO USE;825
17.6.6;5 ACKNOWLEDGEMENTS;825
17.6.7;6 REFERENCES;825
17.7;Chapter
175. A Neural Network for Hyphenation;826
17.7.1;Abstract;826
17.7.2;1 Introduction;826
17.7.3;2 Simulations;826
17.7.4;3 Related Research;828
17.7.5;4 Connectionist versus Symbolic Pattern Matching;828
17.7.6;5 Conclusion;829
17.7.7;6 References;829
17.7.8;6 References;829
17.8;Chapte 176. Temporal Sparse Distributed Memory: Identifying Temporal Patterns via
Homeomorphic Contractions of Memory;830
17.8.1;Abstract;830
17.8.2;Introduction;830
17.8.3;Background and Notation;830
17.8.4;Notation;831
17.8.5;Storage and Pattern Identification (continuous addresses);831
17.8.6;Discrete Addressing;832
17.8.7;Bibliography;833
17.9;Chapter
177. Communication Load Reduction for Neural Network Implementations on Message Passing Multicomputers^;834
17.9.1;Abstract;834
17.9.2;Introduction;834
17.9.3;Method;835
17.9.4;Example Graphs;835
17.9.5;Results;836
17.9.6;Conclusions;837
17.9.7;References;837
17.10;Chapter
178. Teaching a Neural Network to Play GO-MOKU;838
17.10.1;Abstract;838
17.10.2;1. INTRODUCTION;838
17.10.3;2. NETWORK ARCHITECTURE;839
17.10.4;3. IMPLEMENTATION AND PERFORMANCE;841
17.10.5;4. REFERENCES;841
17.11;Chapter
179. The Neural Composer: A Network for Musical Applications;842
17.11.1;Abstract;842
17.11.2;1. INTRODUCTION;842
17.11.3;2. A NEW SEQUENTIAL NETWORK MODEL;842
17.11.4;3. IMPLEMENTATION AND PERFORMANCE;845
17.11.5;REFERENCES;845
17.12;Chapter 180. Quadratic load flow calculation in electric power systems
using a Hopfield model;846
17.12.1;Abstract;846
17.12.2;1. INTRODUCTION;846
17.12.3;2. QUADRATIC LOAD FLOW EQUATION;846
17.12.4;2.QP-BASED LOAD FLOW CALCULATION;847
17.12.5;3. APPLICATION OF A HOPFIELD MODEL TO LOAD FLOW CALCULATION;848
17.12.6;4. CONCLUSION;849
17.12.7;5. REFERENCES;849
17.13;Chapter
181. Object Recognition in Traffic Scenes by Neural Networks;850
17.13.1;Abstract;850
17.13.2;1. BASIC SYSTEM STRUCTURE;850
17.13.3;2. TOWARDS A HYBRID SYSTEM;850
17.13.4;3. VEHICLE-SANN;851
17.13.5;4. TRAINING FILES;852
17.13.6;5. EXPERIMENTS AND RESULTS;852
17.13.7;6. SUMMARY;853
17.13.8;7. Literatur;853
17.14;Chapater
182. A unified theory of edge-preserving smoothing;854
17.14.1;Abstract;854
17.14.2;1. INTRODUCTION;854
17.14.3;2. A NEW INTERPRETATION OF MFA;855
17.14.4;3. SCALE SPACE AND ANNEALING;858
17.14.5;4. RELATIONSHIP TO NEURAL NETWORKS;859
17.14.6;5. CONCLUSION: HOW TO DO IMAGE SMOOTHING;859
17.14.7;6. REFERENCES;861
17.15;Chapter 183. NES: a neural shell for diagnostic expert
systems;864
17.15.1;Abstract;864
17.15.2;1 Motivations;864
17.15.3;2 NES;865
17.15.4;3 Limitations and extensions;869
17.15.5;4 An application;869
17.15.6;References;871
18;AUTHOR INDEX;872